Bug fix to particle flow thinning code

As discussed here:

https://groups.cern.ch/group/atlas-cp-jetetmiss-software-and-validation/Lists/Archive/Flat.aspx?RootFolder=%2Fgroup%2Fatlas%2Dcp%2Djetetmiss%2Dsoftware%2Dand%2Dvalidation%2FLists%2FArchive%2FNPFO%20issue%20in%20zero%20bias%20data%203073&FolderCTID=0x01200200CDF4B276B1C08B43B9321B8EB49890F7#{9337961A-B11C-4D19-8A7A-8D7FDD12DE06}

a bug was found by @markovit whereby the thinning of neutral particle flow objects had not taken place.

Running a data22 test in the latest 23.0 I confirmed this was the case and confirmed that the fix in this MR solves the issue. Thus this MR violates FT0V because it removes neutral particle flow FlowElement with energy < 0 from the AOD.

If we have already produced AOD that will be used in conjunction with future AOD, care might be needed because the jets would presumably also change (assuming the jet finding software does not apply its own cut that input objects should have energy >= 0 - perhaps @delsart knows?). Flagging to @mswiatlo ,@schaarsc ,@camacho ,@mleblanc

Cheers,

MArk